⚡ electrical engineering

Fast 5G Signal Acquisition by Using Non-Uniform Sampling

This paper proposes a deterministic non-uniform (multi-coset) sampling framework for 5G signal acquisition that treats synchronization as a parametric inference problem rather than waveform reconstruction, achieving significant reductions in mean acquisition time (2.8x to 34.2x) while quantifying the associated estimation penalties.

Original authors: Alejandro Gonzalez Garrido, Carla Amatetti

Published 2026-05-19

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Alejandro Gonzalez Garrido, Carla Amatetti

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to find a specific friend in a massive, crowded stadium. Your friend is wearing a bright, recognizable hat (a "known pilot" or signal) and is moving around (Doppler shift) or standing at a specific distance (delay).

The Old Way (Uniform Sampling):
Traditionally, to find your friend, you would hire a team of 100 people to stand in a perfect grid across the entire stadium. Every single person would shout out what they see, every second. This gives you a perfect, high-definition picture of the whole stadium, allowing you to reconstruct exactly where everyone is. However, this takes a lot of time, energy, and money because you are collecting data about empty seats and random strangers, not just your friend.

The New Way (Non-Uniform Sampling):
This paper proposes a smarter, faster way to find your friend without needing that massive team. Instead of a perfect grid, you send out a smaller team of people standing in a specific, pre-planned pattern (a "multi-coset" pattern). They don't stand everywhere; they stand only in the spots most likely to help you identify your friend's hat and movement.

Here is the breakdown of how this works, using simple analogies:

1. The Goal: Finding a Needle, Not Rebuilding the Haystack

The authors argue that in communication systems (like 5G), the receiver doesn't actually need to "rebuild" the entire radio wave (the haystack). It only needs to find the "needle" (the synchronization signal) and figure out where it is and how fast it's moving.

Analogy: If you are looking for a specific song on a radio, you don't need to record the entire broadcast to know when your song starts. You just need to listen for the specific melody. This paper suggests we can skip recording the silence and the other songs entirely.

2. The Strategy: The "Multi-Coset" Pattern

The paper uses a method called Multi-Coset Sampling.

The Analogy: Imagine a clock face with 100 numbers. Instead of checking every single number (0, 1, 2, 3...), you decide to only check specific numbers, like 2, 3, 4, 5, and then skip to 10, 11, 12, 13. You repeat this pattern.
Why it works: Even though you are missing most of the numbers, the specific pattern you chose is mathematically designed so that you can still figure out exactly where your friend is standing and how fast they are moving. It's like solving a puzzle where you only have a few specific pieces, but those pieces are the most important ones.

3. The "Offline" Design: Planning the Perfect Team

Before the receiver even turns on, the authors run a massive simulation (an "offline exhaustive search") to find the best pattern of numbers to check.

The Analogy: Before the game starts, a coach simulates thousands of scenarios to decide exactly which players should stand where. They want a pattern that:
1. Makes your friend stand out clearly (a sharp "peak").
2. Ensures that no matter where your friend is standing in the stadium, the team has an equal chance of spotting them (uniform energy coverage).
The result is a custom-made "sampling map" that is much more efficient than a random guess.

4. The Results: Speed vs. Precision

The paper tested this on 5G signals (specifically the PSS/SSS signals used to connect phones to towers).

The Speed Boost: By using this "skip-and-check" method, the time it takes to find the signal dropped dramatically.
- In some cases, it was 2.8 times faster.
- In the most aggressive tests, it was 34.2 times faster.
The Trade-off: Because they aren't looking at every sample, there is a tiny bit of "blur" in the final measurement.
- Analogy: It's like taking a photo with fewer pixels. You can still clearly see who the person is and where they are, but the image isn't quite as sharp as a 4K photo.
- The paper shows that for the purpose of synchronization (just connecting the phone to the network), this tiny loss in sharpness is a small price to pay for the massive gain in speed.

Summary

This paper presents a framework for fast 5G signal acquisition. It argues that instead of trying to perfectly reconstruct a radio wave (which is slow and expensive), we should treat signal acquisition as a "treasure hunt." By using a smart, pre-calculated pattern to sample only the most useful parts of the signal, we can find the connection point up to 34 times faster than traditional methods, with only a negligible loss in accuracy.

Key Takeaway: You don't need to read every word in a book to find the chapter you need; you just need a smart index. This paper builds that smart index for 5G signals.

Technical Summary: Fast 5G Signal Acquisition by Using Non-Uniform Sampling

Problem Statement
Conventional digital receiver design is rooted in the Nyquist–Shannon sampling theorem, which prioritizes the faithful reconstruction of the entire waveform from uniform samples. However, in signal acquisition and synchronization tasks, the objective is not full waveform recovery but rather parametric inference: determining unknown propagation delay, Doppler shift, and signal amplitude to detect the presence of a known pilot or synchronization sequence. The paper argues that adhering to Nyquist-rate sampling for these specific inference tasks imposes unnecessary processing burdens. The challenge lies in designing a sampling architecture that reduces the data rate and correlator complexity while preserving the statistical properties required for reliable detection and estimation in the delay-Doppler space.

Methodology
The proposed framework utilizes deterministic non-uniform sampling, specifically multi-coset sampling, to address the acquisition problem.

Signal Model: The received signal is modeled as a known pilot sequence affected by unknown delay ( $\tau$ ), Doppler shift ( $\nu$ ), and complex amplitude, corrupted by additive noise. The acquisition task is reformulated as a Generalized Likelihood Ratio Test (GLRT) over a known pilot family.
Multi-Coset Architecture: Instead of arbitrary irregular sampling, the system selects a fixed subset of $K$ sample positions within every block of $L$ Nyquist instants. This creates a periodic non-uniform pattern that can be implemented via synchronized low-rate branches. The average sampling rate is reduced by a factor of $K/L$ .
Compressed-Domain Processing: The framework operates directly on the retained samples. The receiver computes a reduced correlator using only the selected indices, avoiding the reconstruction of the full Nyquist-rate signal. The acquisition metric is normalized by the retained pilot energy to prevent bias toward specific delay hypotheses.
Offline Coset Design: A critical component of the methodology is an offline exhaustive search procedure to select the optimal coset pattern ( $C$ $C$ ) for a given $(L, K)$ $(L, K)$ ratio. The selection criterion is not generic signal reconstruction but a cost function specifically designed for acquisition performance. This cost function jointly minimizes:
1. Peak Isolation: The ratio of the highest sidelobe to the main peak in the delay-Doppler acquisition surface.
2. Energy Balance: The uniformity of retained pilot energy across the entire delay search interval.
  The algorithm evaluates all admissible patterns for a fixed period and selects the one that minimizes the combined cost, ensuring the reduced correlator maintains a dominant peak and unbiased delay coverage.

Key Contributions

Reformulation of Acquisition: The paper shifts the perspective of acquisition from waveform reconstruction to parametric inference, demonstrating that sampling architectures can be optimized to preserve the ambiguity surface rather than the full signal.
Reduced Correlator Structures: It establishes a direct link between multi-coset sampling and reduced correlator structures that operate on compressed measurements, significantly lowering computational complexity.
Task-Aware Pattern Selection: It introduces a specific design procedure for selecting sampling patterns that optimize delay-Doppler estimation performance, rather than relying on generic compressive sensing criteria.

Results
The framework was evaluated using 5G New Radio (NR) synchronization signals (PSS/SSS) under a worst-case Doppler scenario ( $\pm 20$ kHz). The evaluation focused on the trade-off between Mean Acquisition Time (MTA) and estimation accuracy (Root Mean Square Error - RMSE).

Acquisition Time Reduction: The results demonstrate substantial gains in processing speed. Compared to uniform sampling, the proposed method achieved MTA reductions ranging from 2.8 $\times$ to 34.2 $\times$ , depending on the compression ratio ( $K/L$ ). For instance, a configuration with $L=32, K=1$ achieved a 34.2 $\times$ speedup.
Estimation Penalty: The paper quantifies the estimation penalty introduced by aggressive sample reduction. Figures 1 and 2 (Delay and Doppler RMSE) show that while RMSE increases as the sampling ratio decreases, the degradation is manageable relative to the massive gains in acquisition speed.
Trade-off: The results confirm a clear complexity-performance trade-off, where higher compression ratios yield faster acquisition times at the cost of slightly increased estimation error.

Significance and Claims
The paper claims that multi-coset sampling is a natural candidate for synchronization-oriented receivers driven by known pilots or preambles. Its primary significance lies in demonstrating that acquisition capability can be maintained with significantly reduced sampling and correlator complexity without first recovering the full Nyquist-rate signal.

The authors emphasize that for synchronization tasks, where the unknown parameter space is small compared to the signal space, a well-designed multi-coset front-end offers a "meaningful reduction in acquisition time with little degradation on the quality of the estimates." The work validates that the design objective for such receivers should be the preservation of acquisition statistics in the delay-Doppler space, rather than waveform reconstruction. The paper concludes by noting that future work should address multipath effects and experimental validation with live 5G signals.